The fastest way to check whether figures in a paper may have been manipulated is to combine a close visual inspection with forensic image analysis tools, and then cross-check the raw data or methodology whenever possible.
Manipulated figures are a serious concern in academic publishing, and while no single check is foolproof, a systematic approach will catch most red flags. Here is a practical workflow for evaluating image integrity.
1. Start with a careful visual inspection
Before reaching for software, train your eye to spot common tampering signs:
- Inconsistent backgrounds: Look for abrupt changes in brightness, texture, or color within a single image panel (e.g., a Western blot lane that looks “cut and pasted”).
- Repeated patterns: Identical bands, cells, or noise patterns in different parts of an image often indicate cloning or copy-paste manipulation.
- Abrupt edges or misaligned seams: Check where panels are joined; unnatural borders or mismatched resolution are classic red flags.
- Overly perfect data: If every replicate looks nearly identical with zero biological variation, be suspicious.
2. Use forensic image analysis software
Several free and commercial tools can detect duplication and splicing at the pixel level:
- ImageJ with forensic plugins (e.g., “Error Level Analysis” or “Clone Detector”)—free and widely used.
- FotoForensics—a web-based tool that runs error level analysis and noise analysis on uploaded images.
- OpenScience tools like Imaging Integrity or Forensically for quick checks on JPEG compression artifacts.
These tools highlight areas where digital editing has altered the underlying pixel structure, which the human eye often misses.
3. Check for internal consistency and metadata
- File metadata: Right-click the image file and inspect EXIF data. If the file was created in Photoshop or GIMP rather than a microscope or camera, that is a warning sign.
- Scale bars and resolution: Manipulated images often have inconsistent scale bars or lose resolution in edited regions.
- Compare with supplementary files: Many papers hide duplicate images in supplementary materials. If the same figure appears with different contrast or cropping, that is a red flag.
4. Cross-reference with the raw data or methods
If the paper provides raw data (e.g., in repositories like Figshare or Zenodo), compare the published figure against the original. Also, read the methods section carefully—if the authors claim a specific imaging technique, the image should reflect that technique’s typical noise and resolution.
5. Use AI-assisted tools for large-scale screening
If you are reviewing many papers at once, manual inspection becomes impractical. WisPaper’s Scholar QA can help you query a paper’s figures and metadata in context, but for deep forensic screening, consider dedicated integrity-checking platforms used by journal editors (e.g., Proofig or ImageTwin). Scholar QA can help you query a paper’s figures and metadata in context, but for deep forensic screening, consider dedicated integrity-checking platforms used by journal editors (e.g., Proofig or ImageTwin).
When to be suspicious
- Figures that are too clean or too perfect.
- Duplicate images across different figures or papers.
- Unexplained changes in background noise between panels.
- A history of corrections or retractions from the same author group.
Remember: finding a red flag does not automatically mean fraud—it could be a honest mistake or poor image processing. But if you find multiple inconsistencies, it is worth flagging to a journal editor or your supervisor. Combining visual checks, forensic tools, and raw data verification gives you the best chance of catching manipulation before you build your own research on shaky foundations.
